x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

When using eBPF to probe CS CALL instructions within a function,
a crash can be triggered.

The eBPF tool probes offset 257 of the __hrtimer_run_queues()
function:

<__hrtimer_run_queues+249>:  nopl   0x0(%rax,%rax,1)
<__hrtimer_run_queues+254>:  mov    %r14,%rdi
<__hrtimer_run_queues+257>:  cs call <__x86_indirect_thunk_r12>
<__hrtimer_run_queues+263>:  mov    %eax,%r12d
<__hrtimer_run_queues+266>:  xchg   %ax,%ax
<__hrtimer_run_queues+268>:  mov    %r13,%rdi

Which triggers this crash:

  BUG: unable to handle page fault for address: 00000000000f41c9
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP NOPTI
  CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P
  RIP: 0010:__hrtimer_run_queues+0x106/0x230

Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is
at the 6th byte of the above CS CALL instruction. Since the CS CALL
instruction occupies 6 bytes, the exception occurred in the middle of that
call instruction.

The root cause is that when using eBPF tools to probe in the middle of a
function, a kprobe with INT3 is used as the underlying implementation.

During single-step emulation of the original CALL instruction,
int3_emulate_call() assumes that the probed CALL instruction is 5 bytes
long. However, the actual CS-prefixed CALL instruction occupies 6 bytes,
so it constructs an incorrect exception return address. When the CPU
returns from the kprobe handler, the next instruction to be executed is at
the address of the last byte of that CS CALL instruction. Coincidentally,
starting from that address, the CPU fetches and decodes a completely
different instruction, which ultimately triggers a kernel crash.

Fix the issue by using the actual instruction length obtained from
the instruction decoder when constructing the exception return
address, rather than relying on the hardcoded CALL_INSN_SIZE macro.

[ mingo: Refined the changelog ]

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Suggested-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 458 insertions, 0 deletions

diff --git a/fs/minix/dir.c b/fs/minix/dir.c
new file mode 100644
index 000000000..2ca16f849
--- /dev/null
+++ b/fs/minix/dir.c
@@ -0,0 +1,458 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  linux/fs/minix/dir.c
+ *
+ *  Copyright (C) 1991, 1992 Linus Torvalds
+ *
+ *  minix directory handling functions
+ *
+ *  Updated to filesystem version 3 by Daniel Aragones
+ */
+
+#include "minix.h"
+#include <linux/buffer_head.h>
+#include <linux/highmem.h>
+#include <linux/swap.h>
+
+typedef struct minix_dir_entry minix_dirent;
+typedef struct minix3_dir_entry minix3_dirent;
+
+static int minix_readdir(struct file *, struct dir_context *);
+
+const struct file_operations minix_dir_operations = {
+	.llseek		= generic_file_llseek,
+	.read		= generic_read_dir,
+	.iterate_shared	= minix_readdir,
+	.fsync		= simple_fsync,
+};
+
+/*
+ * Return the offset into page `page_nr' of the last valid
+ * byte in that page, plus one.
+ */
+static unsigned
+minix_last_byte(struct inode *inode, unsigned long page_nr)
+{
+	unsigned last_byte = PAGE_SIZE;
+
+	if (page_nr == (inode->i_size >> PAGE_SHIFT))
+		last_byte = inode->i_size & (PAGE_SIZE - 1);
+	return last_byte;
+}
+
+static void dir_commit_chunk(struct folio *folio, loff_t pos, unsigned len)
+{
+	struct address_space *mapping = folio->mapping;
+	struct inode *dir = mapping->host;
+
+	block_write_end(pos, len, len, folio);
+
+	if (pos+len > dir->i_size) {
+		i_size_write(dir, pos+len);
+		mark_inode_dirty(dir);
+	}
+	folio_unlock(folio);
+}
+
+static int minix_handle_dirsync(struct inode *dir)
+{
+	int err;
+
+	err = filemap_write_and_wait(dir->i_mapping);
+	if (!err)
+		err = sync_inode_metadata(dir, 1);
+	return err;
+}
+
+static void *dir_get_folio(struct inode *dir, unsigned long n,
+		struct folio **foliop)
+{
+	struct folio *folio = read_mapping_folio(dir->i_mapping, n, NULL);
+
+	if (IS_ERR(folio))
+		return ERR_CAST(folio);
+	*foliop = folio;
+	return kmap_local_folio(folio, 0);
+}
+
+static inline void *minix_next_entry(void *de, struct minix_sb_info *sbi)
+{
+	return (void*)((char*)de + sbi->s_dirsize);
+}
+
+static int minix_readdir(struct file *file, struct dir_context *ctx)
+{
+	struct inode *inode = file_inode(file);
+	struct super_block *sb = inode->i_sb;
+	struct minix_sb_info *sbi = minix_sb(sb);
+	unsigned chunk_size = sbi->s_dirsize;
+	unsigned long npages = dir_pages(inode);
+	unsigned long pos = ctx->pos;
+	unsigned offset;
+	unsigned long n;
+
+	ctx->pos = pos = ALIGN(pos, chunk_size);
+	if (pos >= inode->i_size)
+		return 0;
+
+	offset = pos & ~PAGE_MASK;
+	n = pos >> PAGE_SHIFT;
+
+	for ( ; n < npages; n++, offset = 0) {
+		char *p, *kaddr, *limit;
+		struct folio *folio;
+
+		kaddr = dir_get_folio(inode, n, &folio);
+		if (IS_ERR(kaddr))
+			continue;
+		p = kaddr+offset;
+		limit = kaddr + minix_last_byte(inode, n) - chunk_size;
+		for ( ; p <= limit; p = minix_next_entry(p, sbi)) {
+			const char *name;
+			__u32 inumber;
+			if (sbi->s_version == MINIX_V3) {
+				minix3_dirent *de3 = (minix3_dirent *)p;
+				name = de3->name;
+				inumber = de3->inode;
+	 		} else {
+				minix_dirent *de = (minix_dirent *)p;
+				name = de->name;
+				inumber = de->inode;
+			}
+			if (inumber) {
+				unsigned l = strnlen(name, sbi->s_namelen);
+				if (!dir_emit(ctx, name, l,
+					      inumber, DT_UNKNOWN)) {
+					folio_release_kmap(folio, p);
+					return 0;
+				}
+			}
+			ctx->pos += chunk_size;
+		}
+		folio_release_kmap(folio, kaddr);
+	}
+	return 0;
+}
+
+static inline int namecompare(int len, int maxlen,
+	const char * name, const char * buffer)
+{
+	if (len < maxlen && buffer[len])
+		return 0;
+	return !memcmp(name, buffer, len);
+}
+
+/*
+ *	minix_find_entry()
+ *
+ * finds an entry in the specified directory with the wanted name.
+ * It does NOT read the inode of the
+ * entry - you'll have to do that yourself if you want to.
+ * 
+ * On Success folio_release_kmap() should be called on *foliop.
+ */
+minix_dirent *minix_find_entry(struct dentry *dentry, struct folio **foliop)
+{
+	const char * name = dentry->d_name.name;
+	int namelen = dentry->d_name.len;
+	struct inode * dir = d_inode(dentry->d_parent);
+	struct super_block * sb = dir->i_sb;
+	struct minix_sb_info * sbi = minix_sb(sb);
+	unsigned long n;
+	unsigned long npages = dir_pages(dir);
+	char *p;
+
+	char *namx;
+	__u32 inumber;
+
+	for (n = 0; n < npages; n++) {
+		char *kaddr, *limit;
+
+		kaddr = dir_get_folio(dir, n, foliop);
+		if (IS_ERR(kaddr))
+			continue;
+
+		limit = kaddr + minix_last_byte(dir, n) - sbi->s_dirsize;
+		for (p = kaddr; p <= limit; p = minix_next_entry(p, sbi)) {
+			if (sbi->s_version == MINIX_V3) {
+				minix3_dirent *de3 = (minix3_dirent *)p;
+				namx = de3->name;
+				inumber = de3->inode;
+ 			} else {
+				minix_dirent *de = (minix_dirent *)p;
+				namx = de->name;
+				inumber = de->inode;
+			}
+			if (!inumber)
+				continue;
+			if (namecompare(namelen, sbi->s_namelen, name, namx))
+				goto found;
+		}
+		folio_release_kmap(*foliop, kaddr);
+	}
+	return NULL;
+
+found:
+	return (minix_dirent *)p;
+}
+
+int minix_add_link(struct dentry *dentry, struct inode *inode)
+{
+	struct inode *dir = d_inode(dentry->d_parent);
+	const char * name = dentry->d_name.name;
+	int namelen = dentry->d_name.len;
+	struct super_block * sb = dir->i_sb;
+	struct minix_sb_info * sbi = minix_sb(sb);
+	struct folio *folio = NULL;
+	unsigned long npages = dir_pages(dir);
+	unsigned long n;
+	char *kaddr, *p;
+	minix_dirent *de;
+	minix3_dirent *de3;
+	loff_t pos;
+	int err;
+	char *namx = NULL;
+	__u32 inumber;
+
+	/*
+	 * We take care of directory expansion in the same loop
+	 * This code plays outside i_size, so it locks the page
+	 * to protect that region.
+	 */
+	for (n = 0; n <= npages; n++) {
+		char *limit, *dir_end;
+
+		kaddr = dir_get_folio(dir, n, &folio);
+		if (IS_ERR(kaddr))
+			return PTR_ERR(kaddr);
+		folio_lock(folio);
+		dir_end = kaddr + minix_last_byte(dir, n);
+		limit = kaddr + PAGE_SIZE - sbi->s_dirsize;
+		for (p = kaddr; p <= limit; p = minix_next_entry(p, sbi)) {
+			de = (minix_dirent *)p;
+			de3 = (minix3_dirent *)p;
+			if (sbi->s_version == MINIX_V3) {
+				namx = de3->name;
+				inumber = de3->inode;
+		 	} else {
+  				namx = de->name;
+				inumber = de->inode;
+			}
+			if (p == dir_end) {
+				/* We hit i_size */
+				if (sbi->s_version == MINIX_V3)
+					de3->inode = 0;
+		 		else
+					de->inode = 0;
+				goto got_it;
+			}
+			if (!inumber)
+				goto got_it;
+			err = -EEXIST;
+			if (namecompare(namelen, sbi->s_namelen, name, namx))
+				goto out_unlock;
+		}
+		folio_unlock(folio);
+		folio_release_kmap(folio, kaddr);
+	}
+	BUG();
+	return -EINVAL;
+
+got_it:
+	pos = folio_pos(folio) + offset_in_folio(folio, p);
+	err = minix_prepare_chunk(folio, pos, sbi->s_dirsize);
+	if (err)
+		goto out_unlock;
+	memcpy (namx, name, namelen);
+	if (sbi->s_version == MINIX_V3) {
+		memset (namx + namelen, 0, sbi->s_dirsize - namelen - 4);
+		de3->inode = inode->i_ino;
+	} else {
+		memset (namx + namelen, 0, sbi->s_dirsize - namelen - 2);
+		de->inode = inode->i_ino;
+	}
+	dir_commit_chunk(folio, pos, sbi->s_dirsize);
+	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
+	mark_inode_dirty(dir);
+	err = minix_handle_dirsync(dir);
+out_put:
+	folio_release_kmap(folio, kaddr);
+	return err;
+out_unlock:
+	folio_unlock(folio);
+	goto out_put;
+}
+
+int minix_delete_entry(struct minix_dir_entry *de, struct folio *folio)
+{
+	struct inode *inode = folio->mapping->host;
+	loff_t pos = folio_pos(folio) + offset_in_folio(folio, de);
+	struct minix_sb_info *sbi = minix_sb(inode->i_sb);
+	unsigned len = sbi->s_dirsize;
+	int err;
+
+	folio_lock(folio);
+	err = minix_prepare_chunk(folio, pos, len);
+	if (err) {
+		folio_unlock(folio);
+		return err;
+	}
+	if (sbi->s_version == MINIX_V3)
+		((minix3_dirent *)de)->inode = 0;
+	else
+		de->inode = 0;
+	dir_commit_chunk(folio, pos, len);
+	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
+	mark_inode_dirty(inode);
+	return minix_handle_dirsync(inode);
+}
+
+int minix_make_empty(struct inode *inode, struct inode *dir)
+{
+	struct folio *folio = filemap_grab_folio(inode->i_mapping, 0);
+	struct minix_sb_info *sbi = minix_sb(inode->i_sb);
+	char *kaddr;
+	int err;
+
+	if (IS_ERR(folio))
+		return PTR_ERR(folio);
+	err = minix_prepare_chunk(folio, 0, 2 * sbi->s_dirsize);
+	if (err) {
+		folio_unlock(folio);
+		goto fail;
+	}
+
+	kaddr = kmap_local_folio(folio, 0);
+	memset(kaddr, 0, folio_size(folio));
+
+	if (sbi->s_version == MINIX_V3) {
+		minix3_dirent *de3 = (minix3_dirent *)kaddr;
+
+		de3->inode = inode->i_ino;
+		strcpy(de3->name, ".");
+		de3 = minix_next_entry(de3, sbi);
+		de3->inode = dir->i_ino;
+		strcpy(de3->name, "..");
+	} else {
+		minix_dirent *de = (minix_dirent *)kaddr;
+
+		de->inode = inode->i_ino;
+		strcpy(de->name, ".");
+		de = minix_next_entry(de, sbi);
+		de->inode = dir->i_ino;
+		strcpy(de->name, "..");
+	}
+	kunmap_local(kaddr);
+
+	dir_commit_chunk(folio, 0, 2 * sbi->s_dirsize);
+	err = minix_handle_dirsync(inode);
+fail:
+	folio_put(folio);
+	return err;
+}
+
+/*
+ * routine to check that the specified directory is empty (for rmdir)
+ */
+int minix_empty_dir(struct inode * inode)
+{
+	struct folio *folio = NULL;
+	unsigned long i, npages = dir_pages(inode);
+	struct minix_sb_info *sbi = minix_sb(inode->i_sb);
+	char *name, *kaddr;
+	__u32 inumber;
+
+	for (i = 0; i < npages; i++) {
+		char *p, *limit;
+
+		kaddr = dir_get_folio(inode, i, &folio);
+		if (IS_ERR(kaddr))
+			continue;
+
+		limit = kaddr + minix_last_byte(inode, i) - sbi->s_dirsize;
+		for (p = kaddr; p <= limit; p = minix_next_entry(p, sbi)) {
+			if (sbi->s_version == MINIX_V3) {
+				minix3_dirent *de3 = (minix3_dirent *)p;
+				name = de3->name;
+				inumber = de3->inode;
+			} else {
+				minix_dirent *de = (minix_dirent *)p;
+				name = de->name;
+				inumber = de->inode;
+			}
+
+			if (inumber != 0) {
+				/* check for . and .. */
+				if (name[0] != '.')
+					goto not_empty;
+				if (!name[1]) {
+					if (inumber != inode->i_ino)
+						goto not_empty;
+				} else if (name[1] != '.')
+					goto not_empty;
+				else if (name[2])
+					goto not_empty;
+			}
+		}
+		folio_release_kmap(folio, kaddr);
+	}
+	return 1;
+
+not_empty:
+	folio_release_kmap(folio, kaddr);
+	return 0;
+}
+
+/* Releases the page */
+int minix_set_link(struct minix_dir_entry *de, struct folio *folio,
+		struct inode *inode)
+{
+	struct inode *dir = folio->mapping->host;
+	struct minix_sb_info *sbi = minix_sb(dir->i_sb);
+	loff_t pos = folio_pos(folio) + offset_in_folio(folio, de);
+	int err;
+
+	folio_lock(folio);
+	err = minix_prepare_chunk(folio, pos, sbi->s_dirsize);
+	if (err) {
+		folio_unlock(folio);
+		return err;
+	}
+	if (sbi->s_version == MINIX_V3)
+		((minix3_dirent *)de)->inode = inode->i_ino;
+	else
+		de->inode = inode->i_ino;
+	dir_commit_chunk(folio, pos, sbi->s_dirsize);
+	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
+	mark_inode_dirty(dir);
+	return minix_handle_dirsync(dir);
+}
+
+struct minix_dir_entry *minix_dotdot(struct inode *dir, struct folio **foliop)
+{
+	struct minix_sb_info *sbi = minix_sb(dir->i_sb);
+	struct minix_dir_entry *de = dir_get_folio(dir, 0, foliop);
+
+	if (!IS_ERR(de))
+		return minix_next_entry(de, sbi);
+	return NULL;
+}
+
+ino_t minix_inode_by_name(struct dentry *dentry)
+{
+	struct folio *folio;
+	struct minix_dir_entry *de = minix_find_entry(dentry, &folio);
+	ino_t res = 0;
+
+	if (de) {
+		struct inode *inode = folio->mapping->host;
+		struct minix_sb_info *sbi = minix_sb(inode->i_sb);
+
+		if (sbi->s_version == MINIX_V3)
+			res = ((minix3_dirent *) de)->inode;
+		else
+			res = de->inode;
+		folio_release_kmap(folio, de);
+	}
+	return res;
+}